Why Diversification Reduces Risk

A portfolio split across assets that do not move in lockstep can carry lower overall risk than any single holding inside it, without giving up any expected return. That outcome is not a sales pitch or an accounting trick. It follows directly from how variance behaves mathematically when independent, or partially independent, sources of uncertainty are combined. The relationship was formalized by the economist Harry Markowitz in his 1952 paper "Portfolio Selection," published in The Journal of Finance, a contribution that later earned him a share of the 1990 Nobel Memorial Prize in Economic Sciences alongside Merton Miller and William Sharpe. This article explains the mechanism, not a recommendation, and it stays entirely at the level of the mathematics and the documented theory behind it.

What diversification actually is, mathematically

Diversification — spreading capital across assets whose returns are not perfectly correlated, so that losses in one holding can be offset, in whole or in part, by stability or gains in another. That definition matters because "owning a lot of different things" is not the same as diversification if those things all respond to the same underlying forces.

The formal starting point is portfolio variance, the statistical measure of how much a portfolio's returns swing around their average. For a portfolio of two assets, variance is not simply the weighted average of each asset's individual variance. It has three components: the variance contributed by asset one, the variance contributed by asset two, and a covariance term that captures how the two assets move together. Markowitz's 1952 paper laid out this framework explicitly, showing that an investor choosing among assets should evaluate not just each asset's expected return and variance in isolation, but the covariances between every pair of assets in the portfolio (Markowitz, "Portfolio Selection," The Journal of Finance, Vol. 7, No. 1, 1952).

Covariance, in turn, depends on correlation — a standardized statistic ranging from -1 to +1 that describes how closely two assets' returns move together, independent of how volatile each asset is on its own. A correlation of +1 means the two assets move in perfect lockstep. A correlation of -1 means they move in perfect opposition. A correlation of 0 means their movements carry no linear relationship to each other. The covariance term in the portfolio variance formula is the product of each asset's volatility and their correlation coefficient, which means that as correlation falls toward zero or below, that term shrinks or turns negative, pulling total portfolio variance down even though nothing about the two assets' individual risk has changed.

This is the entire mechanism in one sentence: diversification lowers portfolio risk because the covariance term in the variance equation shrinks as correlation falls, not because owning more things is inherently safer. Everything else in this article is a consequence of that one algebraic fact.

Why correlation, not headcount, drives the benefit

A common intuition is that more holdings automatically mean more safety. The math says otherwise. Owning ten stocks that all belong to the same industry, say software companies with correlations to each other in the +0.8 to +0.9 range, provides very little diversification, because those stocks share exposure to the same interest-rate sensitivity, the same hiring cycles, and the same investor sentiment shifts. When one falls on bad macro news, most of the others tend to fall with it.

Contrast that with combining two assets of equal volatility whose correlation is 0 instead of +1. Because the covariance term depends directly on that correlation coefficient, dropping correlation from +1 to 0 in a 50/50 combination of two equally volatile assets removes the entire covariance contribution to variance, leaving only the two individual variance terms scaled down by the square of the portfolio weights. Combining two equal-volatility, zero-correlation assets in equal weights can lower portfolio volatility below the volatility of either asset held alone, and the effect gets stronger still if correlation is negative rather than merely zero. None of this changes the portfolio's expected return, which for a simple weighted combination is just the weighted average of the two assets' expected returns regardless of their correlation. Diversification is one of the few places in finance where a genuine "free lunch" of sorts exists: lower risk at the same expected return, achieved purely through combination, not through picking better assets.

How stocks and bonds illustrate the effect

US Treasury securities and equities have historically shown low, and at times negative, correlation during periods of acute market stress. The mechanism behind that pattern is often described as a "flight to quality," a term used in financial-markets research and by institutions such as the Federal Reserve Bank of St. Louis to describe investors selling riskier assets like stocks and moving capital into safer government debt during downturns, which tends to push bond prices up as stock prices fall. That inverse relationship during stress periods is the specific reason a stock-and-bond combination has historically produced a smoother return path than an all-equity portfolio targeting similar long-run expected return, though the strength and even the direction of that relationship can shift across different macroeconomic regimes, a point this article returns to later.

Markowitz's 1952 paper is widely credited as the founding document of what is now called Modern Portfolio Theory. Before it, investment analysis largely evaluated each asset's risk and return in isolation. Markowitz's contribution was to formalize why the *combination* of assets, specifically their covariances, is what determines a portfolio's true risk, a shift in framing significant enough to help earn him the 1990 Nobel Memorial Prize in Economic Sciences.

The two kinds of risk, and which one diversification actually removes

Portfolio theory divides total risk into two categories, and the distinction is central to understanding both the power and the limits of diversification.

Unsystematic risk (also called idiosyncratic or diversifiable risk) is risk specific to a single company, sector, or asset: a factory fire, a product recall, a lawsuit, a management scandal, or a single country's regulatory change. Because this type of risk arises from causes that are largely independent across companies, combining many such assets allows their idiosyncratic shocks to partially cancel out in aggregate, which is precisely the covariance-shrinking mechanism described above.

Systematic risk (also called market risk or non-diversifiable risk) is risk shared by an entire market or asset class: a recession, a broad interest-rate shock, a geopolitical crisis, or a systemic banking event. Because systematic risk affects nearly all assets within a market simultaneously, it cannot be diversified away by combining more assets from within that same market. This distinction between diversifiable and non-diversifiable risk is foundational to the Capital Asset Pricing Model, developed by William Sharpe in 1964 building on Markowitz's framework, which formalizes the idea that investors are compensated with expected return only for bearing systematic risk, since unsystematic risk can, in principle, be diversified away at no cost (Sharpe, "Capital Asset Prices," The Journal of Finance, Vol. 19, No. 3, 1964).

Can diversification eliminate risk entirely?

No. Diversification only reduces unsystematic risk, the risk specific to an individual company or sector. It cannot eliminate systematic risk, the risk shared by an entire market, because during systemic events correlations across nearly all assets within that market tend to rise toward +1, exactly when investors would most want them to stay low.

Why adding assets has diminishing returns

The first uncorrelated asset added to a one-asset portfolio delivers a large risk reduction. The twentieth asset added delivers very little additional benefit, because most of the diversification effect is captured well before that point is reached.

A useful analogy: think of it like adding insulation to a house in a cold climate. The first layer blocks most of the heat loss, because before it, heat was escaping freely. Each additional layer traps a shrinking increment of the remaining heat, until adding still more insulation barely changes the indoor temperature at all. Academic research on US equity portfolios, including foundational work by Meir Statman and earlier studies cited throughout the finance literature, has found that a portfolio of a few dozen stocks spread across different sectors captures most of the diversifiable risk reduction available within a single equity market; the risk that remains after that point is largely systematic risk that no amount of additional stock selection within that same asset class can remove.

This diminishing-returns pattern is a direct mathematical consequence of the variance formula. As the number of roughly equally-weighted, imperfectly correlated holdings grows, the portfolio's variance converges toward the average covariance among the holdings, and each individual asset's own variance contributes a progressively smaller share of the total. Once that convergence has mostly happened, adding one more holding changes very little.

How many holdings are needed to capture most of the benefit?

There is no single universal number, because it depends on how correlated the specific assets in question are with each other. Research on diversified US equity portfolios has generally found that a portfolio of a few dozen stocks across different sectors captures most of the diversifiable risk available within that one asset class, with each additional holding beyond that point adding progressively less benefit.

Where diversification actually breaks down

Diversification's benefit rests on an assumption: that the historical correlation between two assets is a reasonably stable guide to their future correlation. That assumption does not always hold, and understanding where it fails is as important as understanding the underlying mechanism.

Correlations are not fixed physical constants. They are statistics estimated from historical data, and they can shift meaningfully, especially during liquidity crunches. In a liquidity crunch, investors facing margin calls or redemption demands are sometimes forced to sell whatever holdings can be sold quickly, regardless of that asset's underlying fundamentals or its historical relationship to other assets in the portfolio. That forced, indiscriminate selling can drag down assets that had shown little historical relationship to each other, simply because both happened to be liquid enough to sell under pressure. This dynamic was widely documented across asset classes during the 2008 financial crisis, when correlations across equities, corporate bonds, and even some commodities rose sharply and simultaneously, a pattern researchers and market commentators have since described informally as "in a crisis, all correlations go to one."

International diversification faces a related, slower-moving version of the same problem. Owning equities across US, European, and Asian markets historically provided meaningful risk reduction because national economies moved somewhat independently of one another, shaped by different local business cycles, currencies, and policy regimes. As global trade and capital markets have grown more interconnected since the 1990s, a trend documented in research from institutions such as the International Monetary Fund and the Bank for International Settlements on cross-border financial integration, cross-country equity correlations have generally trended upward over time, which reduces, though does not eliminate, the diversification benefit that simply buying equities in more countries used to provide.

Correlation across asset types: a qualitative comparison

The table below summarizes the general, historically observed correlation relationships between broad asset categories, drawn from the qualitative patterns commonly documented in portfolio-theory literature and central-bank research on market behavior during both normal and stressed periods. These are directional patterns, not fixed coefficients, and the exact strength of any of these relationships shifts across different time periods and macroeconomic regimes.

Asset pair	Typical correlation in calm markets	Typical correlation in market stress	Why the relationship shifts
Domestic equities in the same sector	High positive	High positive, often higher	Shared exposure to the same demand cycles, rate sensitivity, and investor sentiment
Domestic equities across different sectors	Moderate positive	High positive	Broad market factors dominate during systemic shocks, compressing sector-specific differences
Equities and government bonds	Low or negative	Historically negative during "flight to quality," but not always	Investors often rotate from risk assets into safe-haven government debt during downturns
Domestic and international equities	Moderate positive, has risen over recent decades	High positive	Global trade and capital-market integration have linked national business cycles more tightly over time
Equities and commodities	Low to moderate, historically variable	Has risen during liquidity crunches	Forced selling during crises can hit unrelated asset classes simultaneously when investors need cash

The pattern that matters most for diversification: correlations that are low in calm markets tend to rise, sometimes sharply, during systemic stress, which is exactly when a diversified investor is depending on them staying low.

Common misconceptions

A frequent mistake is assuming that owning many individual stocks automatically means being diversified. Twenty stocks concentrated in one sector, software companies for example, share exposure to the same interest-rate sensitivity and the same demand cycles, so in terms of portfolio variance they behave more like one large position than twenty independent ones. The number of tickers on a statement says nothing about the correlation structure underlying them.

A second misconception is that diversification is fundamentally about maximizing the count of holdings. What actually drives the effect, based directly on the Markowitz variance formula, is the correlation structure between those holdings, not how many there are. A portfolio of a handful of assets with genuinely low mutual correlation can be better diversified, in the sense of lower portfolio variance for a given expected return, than a portfolio of many highly correlated assets.

A third error is expecting diversification to protect against every category of loss. It reduces the risk that a single company, sector, or country drags down an entire portfolio. It does not protect against a systematic, market-wide decline that affects nearly every asset within that market at roughly the same time, which the Capital Asset Pricing Model framework treats as a fundamentally different category of risk, one that diversification within that market cannot remove.

A fourth, more subtle misconception is treating historical correlation figures as guaranteed future behavior. Because correlations are statistical estimates from past data rather than fixed properties of the assets themselves, portfolios built on assumed low correlations can experience larger-than-expected drawdowns exactly when correlations shift unexpectedly, which the 2008 correlation-convergence episode illustrates.

What this means for building a portfolio, without picking specific assets

None of the mechanism described here implies a specific allocation, and the mathematics of diversification does not by itself tell an investor which assets to combine or in what proportions, since expected returns, risk tolerance, time horizon, and costs are all separate inputs to that decision that fall outside the scope of portfolio-variance mathematics alone. What the theory does establish, with mathematical certainty rather than as a rule of thumb, is that combining assets whose returns do not move in lockstep can reduce a portfolio's total variance below what any single asset in that combination carries on its own, and that this benefit comes without requiring the investor to sacrifice expected return, since expected return for a portfolio is simply the weighted average of the expected returns of its components regardless of how correlated those components are.

It is also worth being explicit about what the theory does not establish. Modern Portfolio Theory as originally formulated by Markowitz assumes that historical variances, covariances, and expected returns are reasonably good estimates of future ones, an assumption that is frequently violated in practice, particularly during regime shifts and liquidity crises as described above. Later academic work, including behavioral-finance critiques and research on fat-tailed return distributions, has documented ways in which the simplifying assumptions behind the original 1952 framework can understate the risk of extreme, simultaneous losses across supposedly uncorrelated assets. None of that undermines the core covariance mechanism, but it is a documented limitation of applying the theory mechanically without accounting for the instability of correlation estimates during stress periods.

The takeaway

Diversification reduces risk because combining assets with low or negative correlation shrinks the covariance term in the portfolio variance equation, allowing the assets' independent price swings to partially cancel each other out, a relationship Harry Markowitz formalized mathematically in 1952 and for which he later shared the Nobel Memorial Prize in Economic Sciences. The benefit is largest with the first few genuinely uncorrelated assets added to a portfolio and shrinks quickly after that point, following directly from how the variance formula converges as more holdings are combined. Critically, the effect weakens exactly when investors would most want it to hold, during systemic crises when correlations across nearly all assets in a market tend to rise toward one. Understanding correlation and the systematic-versus-unsystematic risk distinction, not simply counting holdings, is the difference between genuine diversification and an illusion of it.

---

Read next:

• Why liquidity matters in markets
• Why emergency funds protect decisions
• Why compound interest grows so fast

FAQ

Q: What is correlation and why does it matter for diversification?

Correlation measures how closely two assets' returns move together, on a scale from -1 to +1. Diversification works best when combining assets with low or negative correlation, because their price swings partially offset each other in the portfolio variance formula, lowering overall volatility without changing expected return.

Q: Who developed the mathematical theory behind diversification?

Economist Harry Markowitz formalized the mathematics in his 1952 paper "Portfolio Selection" in The Journal of Finance, showing that portfolio risk depends on the covariances between assets, not just each asset's individual risk. The work later contributed to his share of the 1990 Nobel Memorial Prize in Economic Sciences.

Q: How many holdings are needed to diversify a portfolio?

There is no fixed number, since it depends on how correlated the specific holdings are with each other. Research on US equity portfolios has generally found that a few dozen stocks spread across different sectors captures most of the diversifiable risk available within that single asset class, with each additional holding adding progressively less benefit.

Q: Can diversification protect against a stock market crash?

Not fully. Diversification reduces risk specific to individual companies or sectors, known as unsystematic risk. During broad, systemic market crises, correlations across most assets within a market tend to rise toward one, reducing the protective effect exactly when investors would most want it.

Q: Why do stocks and bonds diversify well together?

They have historically shown low or negative correlation, particularly during downturns, a pattern often linked to "flight to quality" behavior in which investors sell riskier assets and buy safer government bonds during economic stress, pushing bond prices up while stock prices fall. This relationship is not fixed and can shift across different macroeconomic regimes.

Q: What is the difference between systematic and unsystematic risk?

Unsystematic risk is specific to a single company or sector and can be reduced through diversification. Systematic risk is shared by an entire market, such as a recession or interest-rate shock, and cannot be diversified away within that market, a distinction formalized in the Capital Asset Pricing Model developed by William Sharpe in 1964.